Paper Notes¶
Reading notes and analysis on papers I find interesting.
Object Detection
- DINO Emerging Properties in Self-Supervised Vision Transformers
- Stable DINO Stable self-distillation with no labels
- SAM-DETR++ Semantic-Aligned Matching for DETR Convergence
- VLDet Open-Vocabulary Detection with Vision-Language Pre-Training
Transformer Architecture
- Revisiting [CLS] Token decoupling in Vision Transformers
- Naive Sparse Attention Hardware-aligned trainable sparse attention
Large Language Models
- Opus Optimizer-Aware Dynamic Data Selection for LLMs
- GLM-5 Report From Vibe Coding to Agent Engineering